Mask of values below the S1 noise floor

s1tbx
#1

Hi everybody,
I am using S1 EW images collected over sea ice fields (i.e. S1A_EW_GRDH_1SDH_20170914T114425_20170914T114517_018371_01EEBB_B82B or S1B_EW_GRDM_1SDH_20170913T115144_20170913T115244_007373_00D02B_9C2D). Several times the backscattering is below the noise floor of the instument (that depends on the beam angle). This happens usually over standing water between ice packs.
Is there a method to get a binary “backscattering validity map” from SNAP? Just to know if the backscattering value can be consider real or is the noise of the instrument.

Thanks,
Marco

#2

Would ‘S1 Thermal Noise Removal’ followed by some band math operation (e.g. zero thresholding) give you something useful?
I am not sure which values of noise are annotated in the image metadata or how SNAP’s S1 TNR operation works.

#3

Thanks for the suggestion css. Unfortunately, I have not found a proper way to eliminate the pixel with a SNR<1 in SNAP. So I proceed in this way under Matlab:

  • SAR processing (Apply orbit / Thermal noise removal/ Calibration / Reprojection)
  • as described in (doi: 10.1109/TGRS.2017.2721981 sec. II-C), by using the noise equivalent backscattering curves as a function of the incidence angle I classified the image pixels.

The same curves can be found in the Sentinel-1 annual performance report.